JonathanMelhuish.com

I’ve been helping my friend Simon Hammond develop a social networking site designed specifically for mobile internet devices, called Bodder. Although Bodder has been in development for a number of years, it’s never had a clear marketing strategy and has so far only been used by Simon’s friends. As a service that relies on a solid grasp of social dynamics in order to succeed, that doesn’t really sound like a promising approach - we needed to take Bodder out into the real world, see how it was being used, then build that understanding back into the design.

The trouble with Bodder, like many new technologies, is that its true benefits are not immediately apparent, and hence are rather tricky to explain. Another key challenge is that, like other systems that rely on network effects, a social network isn’t much use until a considerable proportion of your friends are using it. Bodder neatly sidesteps this issue by focussing on groups - before you use Bodder to keep track of your friends, you can use it to stay in touch with other members of the same organisations, or to become connected with others who are in the same location or context.

Photo by Simon Hammond

We decided to boil the latter Bodder concept down to its core and ensure that every message had an audience, right from the word go: in exactly 7 days and on a limited budget, we created our own system to allow people to text a message to a large LED ticker display. It’s a fairly well-established concept, but it gets people over the first hurdle - deciding what they want to shout out to the crowd. As soon as they’ve posted a message, they receive a message informing them that they’ve just become the latest member of Bodder, linking them to a page where they can see all of the latest updates from the crowd… and can explore the rest of the Bodder site. Progressively disclosing Bodder’s functionality should make it easier for people to grasp, and increase the likelihood that the hook bites.

Simon wrote a fairly detailed account of how it all went, so I won’t repeat that here, but instead give my interpretation of the results. Although the technology itself worked well, we picked the wrong location to place it in. We chose a prominent, high location that was visible from a large area near the main stage and on a busy route. However, by choosing a position where we were potentially visible to a large number of different people, primarily those who were only passing by briefly, we automatically detached ourselves from any specific audience. This meant that the author of a message had no idea who he was addressing, if anyone, and destroyed any possibility of having a conversation via the screen.

This is probably a good lesson for any internet business to take onboard. On the Internet, it’s tempting to think of your audience as “the world”, but in most cases, you’ll actually attract quite a specific subset, even if the technology is sufficiently generic that it could really be used by anyone. Twitter, for example, is mainly used by the social media crowd, whereas MySpace has long been the home of independent musicians. In a Web 2.0 world, the community is key - not only do you need to know who you’re addressing, your users need to know who they’re addressing too.

This experiment was very successful in teaching us some hard lessons about social dynamics that we could never have learnt sitting at our computers. With these lessons in mind, we’re now planning on taking the same setup to a smaller indoor event where the screen can be visible to the whole crowd. We hope that by making this key change, we’ll get a much better idea of how technology can initiate and support conversations between strangers in the same location. If we can just get them over that first hurdle, who knows what could happen next.

So we identified that online social networks don’t make it easy to find the news that interests you, and we discussed different ways that we can figure out which social context each of your friends fits into. How can we now design a better user interface for browsing news on social networks?

In my dissertation, I suggested interfaces that group the stories according to the person they relate to, then arrange these people in two dimensions according to:

• their social context (by looking at shared friendships and co-appearance in photos)
• how much interest the user shows in stories about this person (by observing the user’s behaviour)

Here’s a short video I made to show one of these designs:


For more interface designs and details about how they could be implemented, please refer to my dissertation (PDF, 3.5Mb).

In my first post in this series, I suggested that current social networks are hobbled by their oversimplified underlying social model. So what can we do to improve this?

In my research, I proposed five categories of friends:

• a close friend whom you see regularly
• a friend who was close but whom you now don’t see or contact regularly
• family
• a new friend whom you see regularly but don’t know much about
• somebody you don’t know well or meet regularly (face-to-face), but publishes good news

Although they seemed like sensible categories, the respondents to my survey only succeeded in categorising an average of 41% of their friends. In retrospect, I was probably rather naïve in assuming that people’s Facebook friends were people they’d with whom they’d had some meaningful relationship at some point in time. In any case, when asked how interested they were in seeing news about each of these categories, there was significantly lower interest in those not covered by these categories, suggesting that I’d not missed out anyone important.

So perhaps these categories have some value in helping people find the news that’s most interesting to them, but they have a key flaw. Not only is it tedious to try to categorise all your friends (the average respondent has 212), but friends will inevitably move between categories.

Facebook takes a different approach - when you add a new friend, they ask instead how you met. This comes back to the idea of friends existing within a social context, something which can actually be quite successfully inferred automatically by simply grouping people according to shared friendships and co-appearance in photographs.

I suggest that there’s probably a link between how much news you’d like to see about a given person and the social context into which those people fit. For instance, you might be quite interested in what your university friends are up to whilst you whilst you’re at university together, but when you graduate you might prefer just to hear about them occasionally - the 21st-century equivalent of the “christmas letter” some of my parents’ friends write.

Trouble is, online social networks such as Facebook don’t do anything useful with this information. Social context is ignored and all your “friends” news is presented you in one big heap. In my next post, I’ll suggest some ways in which the user interface might be re-designed to help you find the news that’s important to you.

For more survey results and discussion of how the social models underlying social networks might be improved, please refer to my dissertation (PDF, 3.5Mb).